Cognizant Technology Solutions has been granted a patent for a text recognition system that utilizes a trained region encoder and instance encoder to process image files. The system modifies images to create data augmentation entities, generating visual instances and output text in response to user prompts through a self-supervised learning approach. GlobalData’s report on Cognizant Technology Solutions gives a 360-degree view of the company including its patenting strategy. Buy the report here.
Smarter leaders trust GlobalData
According to GlobalData’s company profile on Cognizant Technology Solutions, Neural network-based robotics was a key innovation area identified from patents. Cognizant Technology Solutions's grant share as of June 2024 was 60%. Grant share is based on the ratio of number of grants to total number of patents.
Text recognition system with data augmentation and continual learning
The granted patent US12033408B1 outlines a sophisticated text recognition system that utilizes advanced machine learning techniques to enhance the processing of image files containing alphanumeric characters. The system is designed to obtain an image file and a related prompt that queries a specific region within the image. It employs a trained region encoder, which incorporates an attention-based continual knowledge distillation model, to identify a region of interest. The system then modifies the identified image to create a data augmentation entity, generating visual instances from both the original and modified images. These instances are processed to produce ordered sequences, which are subsequently used to further train the region encoder and generate output for user display.
Additionally, the patent details the methodology for training the region encoder through a dual model approach involving a teacher and a student model. The system generates feature sets from the prompt and image file, which are used to create region proposals and calculate a cross-entropy loss metric for model updates. The process also includes generating a prompt vector to facilitate attention-based region determination and updating the global contextual attention engine for improved accuracy. Various operations, such as rotation and scaling, are employed to create the modified image, ensuring robust data augmentation. The system's architecture allows for the integration of different instance decoders, including transformer models, enhancing its versatility in text recognition applications.
To know more about GlobalData’s detailed insights on Cognizant Technology Solutions, buy the report here.
Data Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.
GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.
GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.